A robotic hand with human-like fingers has solved a Rubik's cube in around three minutes. The machine, guided by artificial intelligence, is the first to have managed the feat without being designed specifically for the purpose and to have taught itself. It is built in a way which means it could be used for other things and learnt through a trial-and-error technique known as reinforcement learning. OpenAI taught AI to control the robotic hand which had been developed by the Shadow Robot Company. The AI controlled robotic hand (pictured above) was able to solve the Rubik's cube in three minutes One of the researchers said that the process starts from the very beginning as the AI has to learn how to move the hand.

"This is an interesting and positive step forward, but it is really important not to exaggerate it," said Ken Goldberg, a professor at the University of California, Berkeley, who explores similar techniques. A robot that can solve a Rubik's Cube is not new. Researchers previously designed machines specifically for the task -- devices that look nothing like a hand -- and they can solve the puzzle in less than a second. But building devices that work like a human hand is a painstaking process in which engineers spend months laying down rules that define each tiny movement. The OpenAI project was an achievement of sorts because its researchers did not program each movement into their robotic hand.

Last week, on the third floor of a small building in San Francisco's Mission District, a woman scrambled the tiles of a Rubik's Cube and placed it in the palm of a robotic hand. The hand began to move, gingerly spinning the tiles with its thumb and four long fingers. Each movement was small, slow and unsteady. But soon, the colors started to align. Four minutes later, with one more twist, it unscrambled the last few tiles, and a cheer went up from a long line of researchers watching nearby.

Many tasks in modern machine learning can be formulated as finding equilibria in \emph{sequential} games. In particular, two-player zero-sum sequential games, also known as minimax optimization, have received growing interest. It is tempting to apply gradient descent to solve minimax optimization given its popularity and success in supervised learning. However, it has been noted that naive application of gradient descent fails to find some local minimax and can converge to non-local-minimax points. In this paper, we propose \emph{Follow-the-Ridge} (FR), a novel algorithm that provably converges to and only converges to local minimax. We show theoretically that the algorithm addresses the notorious rotational behaviour of gradient dynamics, and is compatible with preconditioning and \emph{positive} momentum. Empirically, FR solves toy minimax problems and improves the convergence of GAN training compared to the recent minimax optimization algorithms.

OpenAI, the research company that conducts artificial intelligence research was able to train a pair of neural networks to solve the Rubik's Cube using a robotic hand. In a blog post announcing the achievement, OpenAI said the neural networks were trained in simulation, relying on the OpenAIFive code paired with Automatic Domain Randomization, which is a new technique the firm developed. "Human hands let us solve a wide variety of tasks. For the past 60 years of robotics, hard tasks which humans accomplish with their fixed pair of hands have required designing a custom robot for each task. As an alternative, people have spent many decades trying to use general-purpose robotic hardware, but with limited success due to their high degrees of freedom.,"

Artificial intelligence research organization OpenAI has achieved a new milestone in its quest to build general purpose, self-learning robots. The group's robotics division says Dactyl, its humanoid robotic hand first developed last year, has learned to solve a Rubik's cube one-handed. OpenAI sees the feat as a leap forward both for the dexterity of robotic appendages and its own AI software, which allows Dactyl to learn new tasks using virtual simulations before it is presented with a real, physical challenge to overcome. In a demonstration video showcasing Dactyl's new talent, we can see the robotic hand fumble its way toward a complete cube solve with clumsy yet accurate maneuvers. It takes many minutes, but Dactyl is eventually able to solve the puzzle.

Sign in to report inappropriate content. A full solve of the Rubik's Cube. This video plays at real-time and was not edited in any way.

Sign in to report inappropriate content. Perturbations that we apply to the real robot hand while it solves the Rubik's Cube. All videos play at real-time.

We've trained a pair of neural networks to solve the Rubik's Cube with a human-like robot hand. The neural networks are trained entirely in simulation, using the same reinforcement learning code as OpenAI Five paired with a new technique called Automatic Domain Randomization (ADR). The system can handle situations it never saw during training, such as being prodded by a stuffed giraffe. This shows that reinforcement learning isn't just a tool for virtual tasks, but can solve physical-world problems requiring unprecedented dexterity. Human hands let us solve a wide variety of tasks. For the past 60 years of robotics, hard tasks which humans accomplish with their fixed pair of hands have required designing a custom robot for each task.